High speed, high voltage amplifier for cap. load

whoops, that circuit looks funny.

. | A1 . |__________|\\____ T1 ___________| E1+E4 . | | |/ | # | | | . | inv #||# 330 | (Set 1) . | | A2 #||# 10W | . | |___|\\____| #___|_____________________| E2+E3 . | |/ | _|_ | . | AD815 | 1nF_|_ 0.1uF -,- . | dc1 -,- | . | |___| E1+E4 gnd . | ac sig 100k | | . | bus ,--/\\/--' (Set 2) . | | . | |_________________________| E2+E3 . | | _|_ | . | dc2 0.1uF -,- | Gnd

Funny? No, it looks OK to me. Thanks for the detailed description with links and drawings, it was very entertaining! And everything you've described looks OK. The capacitive coupling from one line carrying RF to another carrying RF gave me pause, until I realized they're both carrying the same RF voltage, even if derived from a different amplifier. So that's OK. The capacitance to ground is a necessary evil, and what we've worked to address with enough muscle.

As for that muscle, I'm cautiously-hopeful that the AD815 amplifiers will be up to the task, although John Jardine's comment about their figure 15 says we should ask someone at Analog Devices, or take some measurements, or both.

Amplifiers that can deliver sizable current at high slew rates and high frequencies and with >30V supplies are hard to find, and often become discontinued (witness the nice types formerly from Elantec) without replacements, so I viewed the AD815 as a keeper and a part I should add to our shelves. The SIP package with heatsink-tab Analog offered at first (I have a copy of the older rev-b datasheet) is by far preferable to the 24-lead thermally-enhanced SOIC package they're continuing to sell. So I therefore went looking for any old stock, found some, and purchased 25 pieces ($10) of the staggered-lead pkg, on a lifetime-buy basis. I can send you the source info later from work.

When they come we can check out that figure 15 business. Perhaps in my copious free time :>) I can also design a transformer so we can test the whole circuit to be sure it's up to snuff. If we do this quickly you can decide if you want to get your own lifetime-buy of the discontinued AD815 SIP-tab staggered-lead parts before they disappear.

Well, I ordered some AD815AY amps from Rochester, so I figure I can hook them up and test to see if the large signal freq. response is as it says. I figure that if the response is as it says, that I can build a transformer that gives enough voltage gain at 10 MHz, and then when I want to run lower, I just lower the source voltage (easy to do, since everything is run by cpu).

why is that? They seem like useful objects... though maybe only to me.

That would be great help, since I've never designed one before. I was thinking that since I'm going to be making of order 10 or more of these things, that the transformers should be as uniform as possible over the set -- so, I was thinking of just buying them from some place like coilcraft. Of course, I'm also open to winding my own.

What exactly is a lifetime-buy? Just a bulk order? I'm pretty sure my prof. would be interested in that if this works.

Also, one last out-of-left-field approach. So, we have the following earth-shattering (esp. the A150) amps sitting around lab

ENI 240L

which all like 50Ohm output. These are the amps that I was considering using in the circuits at the beginning of our converasations. Soo, here's the hairbrained idea. What if I do the following (only showing the RF part)

. . |\\ _______ . +-----|/------| Tuner |------| . | Amp +-------+ | . | +--| . Vac . | . Gnd

where Tuner is a HAM radio antenna tuner (e.g. a'la

By the way, any of you guys are welcome to come visit our lab over here at Stanford in the physics department. Not too much to see right now, since we're moving from the basement to the 1st floor (yeah lab upgrade!), though most of the time, we have some other ion traps up and running, so you can come see a single atom, and a pretty kick ass blue laser (here's a slide from a talk I gave awhile ago with a CCD image of a single ion:

Do you happen to know any part numbers for some crt cathode drivers?

These are easy to wind because they don't have very many turns.

That's the term for your decision about how many you'll need to last you the rest of your life, the project's life, the product's life, or whatever. You put those on the shelf and stop worrying about whether they'll be discontinued, replaced, or whatever.

I can't advise you on the use of one of those tuners, except to say this means you can only use one frequency at a time, I assume. Or is the question, how often and fast do you change frequency and how fast does this autotuner tune?

I can advise you on the use of high-power wideband 50-ohm RF amps in your application: it might be reasonable. First, consider, 25 watts on a 50-ohm terminated line is 35.3Vrms = 100Vpp, which is exactly your required voltage. You'd put the power terminator as close as possible to the electrodes. The main problem is your huge 400pF capacitance, which with a 50-ohm source (we assume the RF amp isn't back terminated) places you down -3dB at 8.33MHz.

But 8.3MHz is close to 10MHz and this gives us hope that a simple series peaking inductance can deliver the signal unabated to the electrodes up to 10MHz. For example, a 0.63uH series inductance would resonate with 400pF at 10MHz. Your 50-ohm termination can provide the needed Q-damping for the L-C resonance.

Playing with spice to evaluate the best flat response curve in the 4 to 10MHz region, 0.7uH looks good, assuming a pure 50-ohm source. It peaks to +4% at 5MHz and then drops to -18% at 10MHz (compare to -36% at 10MHz with no series inductance). You can experiment with this stuff on your computer.

It'd probably be best to break up your capacitances into groups and feed each one with its own inductor. For example, with your 16 electrode sets, a simple calculation gives us 25pF each, which means 11uH in series at each electrode set. Or eight sets, 50pF + 22uH, etc. You can measure each one, at the 50-ohm terminator, including its coax cable (which just looks like more capacitance), and calculate the best inductor for that set. Perhaps you'll use small tunable inductors, adjusted to flatten the frequency sweep.

If -20% at 10MHz is unacceptable, or you need a flatter frequency response, you'll have to lower the RF node resistance at the coax junction point. You can use a transformer (e.g., a Jerry Sevick transmission-line type to avoid leakage inductance) or a resistive network. Either way, you'll then need more RF power to get 100Vpp.

One word. Capacitance, my good man, capacitance. Did you notice the test capacitance for those drivers? It was 8pF.

How much capacitance to ground did you add to your modeling, for the floating supply chokes or common-mode transformers being driven at 10MHz?

I am finding the 50-ohm wideband power amplifier attractive. But then they cost the big bucks, and even if there's one or two lying around in the lab, one hesitates to tie it up permanently in an application that can be solved with less horsepower. On the other hand, time is money, and as the ready-made basis for a solution it would take less time.

Kapton, by the way, is the same thing as "polyimide", which is a standard insulation material for transformer winding wire. In my UHV days (1e-11 torr pressure range) I've exclusively used that wire instead of the ultra-expensive stuff the specialized vacuum companies try to sell you. Foil-wrapped has some mechanical advantages over laquered though.

robert

hm, sound very attractive, and my advisor loves throwing $$ around... but it's also fun to make circuits work.

well, I just went down to the Sunnyvale HAM radio outlet here in the bay area and bought an antenna tuner >:) Heh heh, going to hook it up to my monster ENI RF amp with a forward-backward power meter to see if it actually works... otherwise, gotta get to work with some of these other ideas (waiting on the AD815 chips till Mon or Tues, though I've got an APEX PA98 right here, so I'll also start playing with that).

Though... this is California, and I am going surfing in a bit, so we'll see how far I get :-)

This was the sort of one I was thinking of,

maybe its still of some interest to you anyway, you might be able to peak it enough, or just use it in place of the ad815s with op transformers.

Incidently out of interest I ran my idea of chained bootstraped op amps through ltspice and it looked surprisingly good although I since found out the model for the LT1210 has grnd referenced nodes wich makes the results less reliable.

Colin =^.^=

antenna tuner was "okay," but not great. It reduced the reflected power at some frequencies, but not at others (doesn't seem to like

Hi, True 8pf isnt much although its still requires .1 amp, also their test ciruit does have quite an involved output ciruit, but considering they are rated for up to 200mhz pixel clock I would think they would be able to cope with considerably higher capacitance at only 10mhz, they have fixed gain so there is no feedback loop to become unstable with higher capacitances. They dont however give much information about what the SOA is although they say that resistive loads can be used for other aplications limited only by thermal considerations.

They are quite cheap and used enmass, and as theres 3 in a package you could distrubute the electrodes as necessary, I would be inclined to think that devices in the same package would have close enough charecteristics to be paralleled fairly easily.

However as they dont have the voltage swing I was thinking they did they seem far less atractive, as they still need either a transformer or peaking/matching circuit, otherwise they would of been a neat solution.

Well I wasnt too sure what capacitance to add, or to what nodes (10pf+ to each op might not be unreasonable), seing as though one might need to put the thing near some metal to heatsink it too, also device inductances might need to be taken into consideration too, the file is in the ltspice grp, I tweaked it a bit since then, however as the op-amp model uses internal gnd nodes I didnt bother to investigate any further. I was mainly interested in seeing what the current distribution was like. might be useful if you find you need a freq range too wide for a transformer.

I would sugest going with Win's ideas anyway as there seems less unknowns, and hes been good enough to work out most things already :) although you might consider the lt1210 op amp as a close alternative (single op amp per package/higher curent)

Colin =^.^=

But don't ditch the wideband power amp just yet, check the comments on a series L in my post, Message-ID: To be clear, I'm talking about adding fixed series inductor(s), with no tuning vs frequency required.

That's a nice idea, but your addition of a 50-ohm source impedance (which I doubt can be assured from a typical RF power amp) gives you a great advantage.

In fact, with an assured 50-ohm source, the 50-ohm termination now presents a 25-ohm impedance to the 400pF. This means that a straight connection to the 400pF with no inductance is down only 14% at 10MHz, and with a 0.135uH inductor in series it's a virtually perfect +0.3, -0.6% response, and thus better than your network.

Apparently Jesse would need to characterize the Zout of his amplifier before calculating his best solution, or do an empirical evaluation. That is, try it.

...

Going further you can start with a 4th order cheb matching filter and then taylor it for just 400pF output cap and best output. The best case would be to have an additional 50R load.

This gives you something like this:

.--------------. 680nH 1uH | ___ | ___ ___ | .-|___|--------------UUU--+--UUU-----+--------. | | 50R | | | | | /+\\ | | .-. | electrode cap | ( ) | --- 50R | | --- | \\-/ | 520p --- Load| | --- 400p | | | | '-' | | | | | | '--------------' | | | HF amplifier === === === GND GND GND

There you have 0/-4% over a 10.5MHz BW.

You can do that too and the components values become 11uH, 16uH, 33p, and

...

I've never used wideband RF amps and have simply assumed they could handle the power. I'm apparently mistaken.

Oops. Overlooked this. But then if the RF amp doesn't cope with the 50R load you're stuck too. Unless designing the filter for a higher impedance load, 100R for ex.

This will give you +0.7%/-3.3% over the bandwidth and -2.5% up to 9.9MHz. Unfortunately the input impedance drops down to 26R at 5MHz and we're almost back to square one. Almost because its phase is better behaved than the series L network (+/- 20 degrees) and is almost resistive.

Sure.

okay, thanks so much for all the help! I got the AD815s in, so I'm taking a look at transformers now as well. Also, I'm taking a closer look at your power amp comments as well... any suggestions for winding a transformer for the AD815's? Should I use ferrite cores?

Jesse

W> >